The Technology Innovation Institute (TII), functioning as the applied research arm of Abu Dhabi’s Advanced Technology Research Council (ATRC), has launched the second iteration of its renowned large language model (LLM), Falcon 2. This release introduces two significant versions: Falcon 2 11B, a more efficient LLM trained on 5.5 trillion tokens with 11 billion parameters, and Falcon 2 11B VLM, notable for its vision-to-language model (VLM) capabilities, facilitating seamless conversion of visual inputs into textual outputs.
An exceptional feature of Falcon 2 11B VLM is its status as TII's inaugural multimodal model, making it the only one currently available in the top-tier market with image-to-text conversion capability, marking a substantial leap in AI innovation. In comparative tests with several prominent AI models, Falcon 2 11B outperforms Meta’s recently launched Llama 3 with 8 billion parameters and matches the performance of Google’s Gemma 7B, as confirmed by independent verification from Hugging Face, a US-based platform hosting objective evaluation tools and global leaderboards for open LLMs.
Significantly, both Falcon 2 11B and 11B VLM are open-source, providing developers worldwide with unrestricted access. Plans include expanding the Falcon 2 next-generation models to include a variety of sizes and integrating advanced machine learning capabilities like 'Mixture of Experts' (MoE) to further enhance their performance. Consistently ranked among the top-tier globally, TII’s AI models, including the new Falcon 2 11B variants, are poised to gain wider market acceptance in the dynamic realm of generative AI.
Falcon 2 11B models, with multilingual capabilities, handle tasks in various languages, enhancing versatility. Falcon 2 11B VLM, a vision-to-language model, interprets images for applications in healthcare, finance, e-commerce, education, and law. It aids document management, digital archiving, and context indexing, and assists visually impaired individuals. These models are efficient on a single GPU, ensuring scalability and easy integration into lighter infrastructures like laptops.
